Morning Edition · Friday, October 23, 2026
Tech
Anthropic Says Claude Formally Verified 123 Open-Source Libraries at the Assembly Level
The machine-checked proofs total about 180 million lines of Rocq, cover 239 fuzzing entry points, and came with patches for bugs that trillions of fuzz runs had missed.
Tech
Major US Banks Block or Restrict Outside AI Agents That Log In With Customers' Credentials
Morgan Stanley is blocking third-party agents and JPMorgan, Bank of America and Citi are restricting them, after banks saw agents take actions that went beyond what customers asked for.
Tech
OpenAI's GPT-6 Models Enter Private Preview on Google Cloud, Completing Its Spread Across the Three Largest US Clouds
Google Cloud customers can now apply their existing spend commitments to GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, and OpenAI models are coming to Google's Antigravity coding tool and Gemini Agent.
Tech
PHP Ships Emergency Patches for Two Critical Remote-Code-Execution Flaws That an OpenAI Audit Helped Find
The two bugs score 9.2 out of 10, may expose more than a million internet-facing servers, and pose the highest risk to Apache servers running mod_php.
Tech
Microsoft Researchers Prove the 31-Year-Old KLS Conjecture With Human-Directed AI Agents
The agents explored about 100 candidate approaches under human direction, while fully autonomous runs from start to finish failed to produce a proof.
Tech
Nvidia Pitches Europe on Sovereign AI Factories in Berlin, Citing Nearly $10 Billion Invested in a Year
Jensen Huang's GTC Berlin keynote centered on physical AI and agent infrastructure, and Nvidia launched Minerva, an AI-driven operations platform for monitoring GPU clusters.
Tech
Researchers Show Facility Power Meters Can Detect Hidden AI Training Runs Without Trusting the Operator
On simulated data center power, a Viterbi tracker caught at least 96% of undisguised training runs at a 1% false-alarm rate, and at least 87% of runs over a day even when the operator tried to hide them.
Tech
Poland's NASK Gains Access to OpenAI's Restricted GPT-5.5 Cyber Model Through a Government Partnership
Poland becomes the third European country in OpenAI's trusted-access program for cyber models, as Anthropic separately widens its program giving vetted security professionals access with fewer safeguards blocking requests.
Geopolitics
Kazakhstan Moves From Plans to Deployment With Hundreds of Nvidia GPUs Running and Vera Rubin Systems Planned
Kazakhtelecom has deployed 400 H200 and 480 B300 GPUs in Astana and Almaty and plans more than 1,000 Vera Rubin GPUs in the Ekibastuz region next year.
Tech
Real2Sem2Real Turns One Human Video Into Thousands of Robot Demonstrations That Match Teleoperation Data
The pipeline generates more than 2,000 training episodes in about two hours from a single monocular RGB recording, and policies trained on that data matched or beat teleoperation baselines on two real robots and four tasks.
World
Google DeepMind Publishes a 10-Meter Global Map Separating Primary Forest From Plantations and Tree Crops
The map's six classes follow the definitions of the UN Food and Agriculture Organization and the EU Deforestation Regulation, so it bears directly on importers of cocoa, coffee, palm oil and rubber.
Tech
"Cognitive Commons" Paper Argues AI Adoption Could Erode the Expertise Needed to Oversee AI
The framework's "Validation Tether" holds that effective AI oversight depends on deep human expertise, the same expertise that rational, firm-by-firm AI adoption may stop producing.